Repo is the least visible large market in finance. It connects the cash of money-market funds and corporate treasuries to the collateral of dealers, banks and hedge funds, and in doing so finances a substantial share of the securities positions in existence. Its basic unit of time is overnight — and almost nobody asks why. The answer is not economics. It is machinery: settlement cycles, custody transfers, eligibility checks, valuation, margin and reconciliation, each slow enough that one night became the shortest interval the system could reliably complete. Tokenisation removes most of that machinery. Intraday repo with maturity specified to the minute is not a forecast; it has been running in institutional production for years.
This essay is about what that changes — not the plumbing, which is by now demonstrated, but what the plumbing does to the price of credit. The argument runs in three steps. Below overnight, time stops being the thing that is priced, and balance-sheet capacity takes its place. The demand for minute-level money comes not from banks — which already borrow intraday from central banks for nothing — but from everyone who cannot. And the end effect of shortening money is not uniformly cheaper credit but a redistribution of credit between bank and market channels, with consequences for every borrower who can only reach one of them.
The reference rate of dollar finance is a repo rate. SOFR — successor to LIBOR, benchmark for a stock of derivatives and loans measured in the hundreds of trillions of dollars — is computed each morning from the previous night's Treasury repo transactions (Federal Reserve Bank of New York, SOFR). It is not a rate at which banks estimate they would lend to one another; it is the observed price of overnight cash against collateral. The entire pricing architecture of dollar finance rests, in other words, on a market most people have never seen — and on a tenor almost nobody has thought to question.
The market itself is simple in function and vast in scale: cash pools — money-market funds, corporate treasuries — lend against securities; dealers finance their inventories and their clients' leverage; several trillion dollars is arranged, collateralised and unwound every day across the bridge between the money market and the securities markets, with the balance sheets of a small number of dealers as the pillars the bridge stands on. Term repo exists — a week, a month, three months — but the atom of the market, the unit from which everything else is built and the unit embedded in the world's reference rate, is one night.
The interesting question is not how large the market is. It is why its basic unit of time is what it is — and what happens to everything built on top of that unit if it stops being the minimum.
The intellectual map of this territory was drawn by Zoltan Pozsar. His work on shadow banking and institutional cash pools described a system in which the binding quantities are not deposits and loans but cash, collateral, funding and dealer balance-sheet capacity — a system of market-based intermediation running parallel to the banking system and, in wholesale terms, larger than it (Pozsar et al., FRBNY Staff Report 458, 2010; Pozsar, IMF WP/11/190, 2011). Institutional cash pools are too large for insured deposits, so they hold repo and money-market instruments instead; the supply of safe collateral, not the supply of reserves, becomes the constraint on private money creation; and dealer balance sheets are the transformers through which one becomes the other.
What that map holds constant is the clock. Every instrument in Pozsar's architecture carries a tenor drawn from the same short menu, and the bottom of the menu is overnight. The architecture was never designed around that floor; it inherited it. Which is precisely why the floor is worth examining.
Pozsar mapped the architecture of short-term money. Tokenisation changes its clock.
There is no theorem that says secured funding must last at least one night. Overnight became the floor because the operational chain beneath a repo transaction — settlement in central securities depositories, custody transfers, collateral eligibility checks, valuation, haircuts, margining, reconciliation between counterparties, and the legal machinery of transfer — takes hours to complete and was built to complete once per day. A funding transaction cannot be shorter than the machinery that executes it. One night was never the demand side's preference. It was the supply side's minimum.
The consequence is that the financial system carries liquidity it does not need for liquidity's sake. Part of every institution's buffer exists because liquidity cannot be obtained exactly when it is needed, in exactly the amount needed, for exactly the period needed. The buffer is a standing tax paid to settlement friction.
But honesty requires the other half of the story. Overnight is not only friction; it is a load-bearing convention. Interest accrues daily. Netting runs at end of day. Risk systems, funding desks and regulatory reports operate on a daily cycle. And the reference rates of the entire financial system — SOFR, €STR, and their predecessors — are overnight rates, because overnight is the tenor at which central banks implement policy. The floor under money's duration is operational in origin, but a great deal of institutional weight now rests on it. Removing it is not a neutral act.
It is important to be precise about what is demonstrated, what is emerging, and what is extrapolation — because the first category is larger than most commentary assumes.
Demonstrated: intraday repo in institutional production. J.P. Morgan's blockchain platform, now Kinexys, executed its first intraday repo transactions in late 2020; BNP Paribas traded on it in May 2022; Santander executed its first programmable intraday repos — euro and dollar legs, executing at a programmed time and redeeming three hours later — in January 2025 (Santander CIB, January 2025). The intraday repo application alone has enabled more than $300 billion in trading volume, and the platform as a whole has processed over $3 trillion in transactions, averaging more than $5 billion daily (J.P. Morgan, April 2026). The 2025 interoperability link between Kinexys and HQLAx was announced in language worth quoting for its casualness: repo traders can exchange cash and collateral intraday "with settlement and maturity times specified to the minute" (J.P. Morgan / HQLAx / Ownera, August 2025). J.P. Morgan's own product page states the pricing convention still more flatly: interest accrues on actual usage, down to the minute (J.P. Morgan, Digital Financing). Broadridge's distributed ledger repo platform processes on the order of $7.5 trillion a month across its repo workflows (Broadridge, June 2026). Clearstream announced an intraday triparty repo service in March 2026 (Finadium, March 2026). And the priority ordering is now documented from the regulator’s side: across the 123 responses to the FCA and Bank of England’s joint Call for Input on wholesale tokenisation, collateral was by far the most frequently cited use case, while 24/7 trading and atomic settlement were rarely mentioned except where they related to collateral (FCA / Bank of England, FS26/1, September 2026).
Emerging: the surrounding stack — tokenised money-market fund shares as mobile collateral, tokenised deposits and wholesale settlement money, the central-bank bridges that connect on-chain settlement to sovereign money. That terrain deserves a survey of its own; for present purposes, what matters is that every element of it exists in at least institutional pilot form.
Extrapolation: funding measured in seconds, and repo as a continuous background process rather than a discrete transaction. That is where this essay is heading — but note what the demonstrated category already establishes. Term granularity at the minute level is not a research agenda. It is a product. The question posed by a forty-seven-second repo is no longer whether it can be done. It is whether it is worth anything.
So run the arithmetic. A dealer needs $500 million for forty-seven seconds — a settlement mismatch, a margin call arriving before an inflow, a momentary gap between two legs of a trade. At a four per cent money-market rate, the interest on that transaction is approximately thirty dollars.
Thirty dollars. The number is the argument. Interest on sub-minute money rounds to zero, which means the funding curve does not extend smoothly below overnight. There is no meaningful term structure between one second and one night, because there is almost no time value to structure. What a sub-overnight market prices is not time. It is everything else: the per-transaction cost of the machinery, the availability of the facility when it is needed, the collateral mobility that makes it possible, and above all the balance sheet on the other side. The natural pricing model is not an interest rate on a term. It is a capacity charge — a fee for the standing option to fund, plus a toll per use.
This has an unfashionable implication for the tidy image of a continuous duration spectrum running from seconds to months. The spectrum does not smooth into a curve. It bifurcates. Above overnight, time is priced, and the term structure works as it always has. Below overnight, time is nearly free and capacity is scarce, and the market that forms there looks less like a lending market than like an infrastructure service — closer in economics to a payment system than to a money market.
Below overnight, time is no longer the thing being priced. Capacity is.
The second unfashionable observation: for banks, the incumbent competitor in sub-day funding is free. The Federal Reserve extends collateralised daylight overdrafts to account holders at a zero fee under its Payment System Risk policy; the Eurosystem provides intraday credit in TARGET against eligible collateral without interest (Federal Reserve PSR policy; ECB TARGET Guideline). A bank with central-bank access and spare eligible collateral already has minute-level liquidity at a price no private market can undercut. Whoever expects banks to pay meaningfully for tokenised intraday funding during central-bank hours has not looked at what they currently pay.
The demand side of the sub-overnight market is therefore specific, and worth naming. First, everyone without an account at the central bank: hedge funds, principal trading firms, money-market funds, the non-bank entities of dealer groups — the population whose share of intermediation has grown for fifteen years and which currently buys its intraday liquidity implicitly, through buffers, credit lines and the goodwill of clearing banks. Second, the hours the central bank does not keep. Tokenised markets settle continuously; central-bank money keeps office hours, and even the bridges now being built between DLT settlement and sovereign money operate on the central bank's calendar. A market that never closes has liquidity needs at times when the free alternative is shut — and although the Bank of England is consulting on extending RTGS and CHAPS settlement hours to near-24/7, that widens the perimeter’s clock, not its membership (FCA / Bank of England, FS26/1, September 2026). Third, the gaps between systems and currencies, where a position is flat by end of day but funded nowhere in between.
Sized honestly, then, the sub-overnight market is not a replacement for repo. It is the extension of intraday liquidity beyond the central-bank perimeter and beyond the central-bank day — the privatisation, at a price, of a facility that banks receive from the state for nothing. That is a narrower market than the rhetoric suggests, and a structurally important one, because it is precisely the non-bank system that has no lender of last resort at any tenor.
Shorten the funding cycle and the same collateral can do more work. An asset that secures one overnight transaction today can, in principle, secure several sequential intraday transactions tomorrow: mobilised, returned, mobilised again. Utilisation rises; less collateral sits trapped in transit or buffer; the settlement-driven share of liquidity reserves shrinks. Tokenisation does not create collateral. It raises its velocity.
That word has history. Manmohan Singh measured the velocity of pledged collateral — the length of the re-use chains through which one security supports multiple layers of funding — at roughly three before 2008, falling sharply afterwards as regulation and counterparty caution shortened the chains (Singh, IMF WP/11/256, 2011). The post-crisis decade deliberately made collateral slower. A technology whose core financial property is making collateral faster is, in this one specific sense, rebuilding what regulation dismantled — with better records and shorter intervals, but the same underlying physics: longer chains mean more interconnection, and interconnection is what turned a mortgage problem into a funding crisis.
Meanwhile the scarce resource migrates. If collateral is continuously mobile, the binding constraint on secured funding becomes the balance sheet willing to intermediate it — leverage capacity, not asset availability. And here sits a quiet measurement problem: the regulatory perimeter is built on end-of-day photographs. Leverage ratios, G-SIB scores and liquidity ratios are computed at reporting snapshots; the Basel framework's intraday liquidity tools are monitoring indicators, not binding requirements (BCBS 248, 2013). A balance sheet that expands violently at 11:04 and is flat by the close is, to the regulatory eye, a balance sheet on which nothing happened. The market is moving to seconds. The perimeter measures in days.
Every funding architecture needs a resting place for cash, and in this one it is the tokenised money-market fund: short-duration government exposure, yield-bearing, and — the operative property — pledgeable. The first live institutional use of tokenised MMF shares was exactly this: BlackRock fund shares transferred to Barclays as derivatives collateral in October 2023, without redeeming a single share (J.P. Morgan TCN, October 2023).
It matters to separate two channels here, because they are routinely conflated. Money funds competing with bank deposits is not new and owes nothing to tokenisation; it is a story that begins with Regulation Q in the 1970s and was most recently rehearsed in the deposit migrations of 2023. What is new is the collateral utility of the fund share. A share that can be pledged in minutes does not need to be redeemed to be spent, which makes it operationally closer to money than any prior fund format — while simultaneously removing part of the classic run mechanics, because a holder who can pledge under stress has an alternative to redemption, and redemption is what forces asset sales. The same property that makes the tokenised MMF a sharper deposit competitor makes it a milder fire-sale engine. Both effects are real, and they do not net to zero in any state of the world; they simply bind in different ones.
The full stack that assembles around this asset — fund share as collateral, repo against it, settlement money beneath it — is a survey of its own. What needs stating here is the consequence: when the deposit substitute is also infrastructure, the question of where short-term savings sit stops being a product question and becomes an architecture question.
The macroeconomic bridge is the one this essay exists to cross. Banks do not lend whatever borrowers demand; funding cost, liquidity, capital and balance-sheet constraints shape the economics of every loan. Change the architecture of short-term funding and you change those constraints — and therefore, eventually, the supply and price of credit.
Two channels pull in opposite directions. Through the first, deposits migrating into pledgeable fund shares make bank funding more contestable at the margin: more wholesale funding, higher marginal cost, pressure on the economics of balance-sheet lending. Through the second, cheaper and more precise secured funding strengthens exactly the market-based channel — dealers, funds, securitisation — that competes with bank lending. The honest answer to whether tokenisation makes credit cheaper or dearer is that it does neither universally. It redistributes credit supply between the bank channel and the market channel.
And redistribution has an address. Borrowers whose assets can reach the market channel — large corporates, anything securitisable, anything that can be collateral — stand on the side where funding is getting cheaper and more continuous. Borrowers who can only be reached by a banker's judgement on a bank's balance sheet — small firms, most households in bank-centric systems — stand on the side whose funding is becoming more contestable. The shortening of money widens the spread between market-reachable credit and bank-dependent credit. No economy is more exposed to that spread than Europe, which combines the most bank-dependent borrowers in the developed world with a policy programme actively promoting the market channel.
Everything above is the efficiency ledger. The other ledger opens with a simple symmetry: rails that mobilise collateral in seconds transmit stress in seconds. Faster funding is also faster withdrawal; automated margining is also automated margin spirals; programmable unwind is pre-programmed pro-cyclicality. The repo dislocations of September 2019 (Anbil, Anderson & Senyuz, FEDS Notes, February 2020) and the dash for cash of March 2020 — including the margin spirals in fixed-income markets documented at the time (FSB, November 2020; Schrimpf, Shin & Sushko, BIS Bulletin No. 2, April 2020) — played out at overnight speed. Which meant there was a night between decisions: an interval in which desks could meet, positions could be reviewed, and a central bank could announce a facility before the next funding round. Compress the funding cycle to minutes and that interval compresses with it.
The overnight pause was never designed as a circuit breaker, but it functions as one. It is when risk committees convene, when humans overrule models, when authorities act. A liquidity architecture with no minimum duration has no built-in moment at which anyone is obliged to stop and look.
The same friction that traps liquidity also slows contagion.
None of this is an argument for keeping friction. It is an argument for pricing its removal honestly: the efficiency gain and the reflexivity cost are the same property, observed in calm and in stress respectively. Smaller buffers, higher collateral velocity and precise funding are what resilience is spent on.
One consequence is flagged here rather than resolved, because it deserves its own essay. Monetary policy is implemented at the overnight tenor. The operational targets are overnight rates; interest on reserves and the deposit facility accrue daily; the corridor assumes that overnight is where the marginal funding decision happens. If the marginal funding decision migrates below one day — into a tenor where the central bank currently gives banks free credit and gives non-banks nothing — the implementation framework faces a fork: extend the corridor into the day and start pricing daylight liquidity, or leave the sub-day segment to a private capacity market and accept that the policy rate anchors a tenor the system is vacating. Neither branch is trivial. Both are, for now, unexamined.
Overnight repo is not disappearing. What it is losing is its status as the floor — the shortest meaningful duration of money, the atom from which the money market is built. Beneath it, a segment is forming in which funding duration approaches the duration of the liquidity need itself, in which time is nearly free and capacity is the scarce good, and in which the participants are precisely those the official liquidity system does not serve.
Follow the causal chain to its end: the clock changes; the scarce good changes from time to capacity; collateral becomes more productive; balance-sheet capacity becomes more valuable; the economics of bank and market intermediation shift; the allocation and price of credit follow. The shortening of money begins as an operational curiosity in the plumbing of repo and ends as a change in who supplies credit to whom, and at what price. Banks are not removed from that end state; they are repositioned. Deposits become less unique. What becomes more valuable is the one thing that cannot be tokenised: a trusted balance sheet, standing behind a promise, at the moment the machines are all trying to unwind at once.
When liquidity has no minimum duration, the last scarcity is the balance sheet that stands behind it.
The forty-seven-second repo is an illustration; its figures are stated assumptions, not market data. This essay was developed with AI assistance (Anthropic's Claude); the analysis, positions and final text are the author's.
Written in a personal capacity. Analytical views only — not legal or investment advice.
Julian Gretzinger — Investor and writer on monetary history, real wealth mechanics, and financial markets. substack.com/@juliangretzinger